Founded in 2017 by former Samsung Electronics and AMD engineer June Paik, Furiosa AI specializes in developing AI inference processors for data center servers.
The Seoul-based company debuted its first AI accelerator chip, also known as "NPU," "Warboy" in 2021, followed by its next-generation processor "RNGD" engineered to handle large language model (LLM) AI in August last year.
Industry experts note that Furiosa AI's second-generation NPU RNGD surpasses market leader Nvidia's GPU L40S and even the high-end version of the U.S. tech giant's H100 in terms of performance-per-watt ratio for inference operations.
The firm has secured about $115 million in funding to date, including a recent $1.5 million investment from venture capital CRIT Ventures in early February.
The potential acquisition is in line with Meta's strategy to decrease its reliance on Nvidia's costly AI chips. The social media behemoth is already partnering with American semiconductor manufacturer Broadcom to develop custom AI processors.
Meta has also recently unveiled plans to invest up to $65 billion in AI infrastructure and data center expansion this year.
Copyright ⓒ Aju Press All rights reserved.
